Amazon is in talks for Decart at about $7 billion. Price the chip story, not the headline.

In brief

Amazon is in talks to acquire Decart, a San Francisco AI startup with Israeli roots, in a deal that could value it at around $7 billion, the Wall Street Journal reported on 11 October 2026, citing people familiar with the matter. Neither company has confirmed an agreement, and Reuters said it could not independently verify the report. Anthropic explored buying Decart in August for about $6 billion and walked away in September after due diligence, according to Reuters and Bloomberg reporting.

What matters for technology leaders is less the price than what Decart would bring into Amazon's stack: software that runs models efficiently across chips from different vendors, and real-time video models that already run on Amazon's Trainium accelerators.

What Decart sells

• Chip-flexible inference software that helps developers move workloads between Nvidia GPUs, AWS Trainium and other accelerators with less re-engineering.
• Real-time generative models: Lucy for video and Oasis for interactive simulated environments.

The Journal reported that AWS gave Decart early access to Trainium3 in 2025, and CTech reports AWS is a strategic customer. Decart's own figures claim large frame-rate gains on Trainium over GPUs; those are vendor claims, not benchmarks ByteNib has verified.

Why $7 billion is a signal, not a close

It is a steep premium. Decart raised $300 million at nearly $4 billion in May, in a round that included Nvidia. Nearly doubling in five months implies a strategic buyer.

The asset has been shopped. Nvidia interest, then Anthropic's diligence and exit, then Amazon. Anthropic's reasons were not disclosed, so assume this deal could still break.

Portability cuts both ways. Tooling that eases switching away from Nvidia helps Amazon pull workloads onto Trainium. It also helps customers keep exit routes. Whether Amazon keeps it open or couples it to Trainium decides which.

What leaders should do next

1. Do not change vendor strategy on a rumour. Keep accelerator roadmaps on their existing review cadence until there is a signed deal.
2. Ask AWS the Trainium question. If you are evaluating Trainium or Bedrock-hosted generative workloads, ask how Decart's stack would be offered and what stays portable off AWS.
3. Price switching cost explicitly. Put a quantified exit cost for accelerator lock-in into the next capacity plan.
4. Watch the diligence outcome. If Amazon closes quickly, ask what it cleared that Anthropic did not.

Sources and scope

• Wall Street Journal, "Amazon in Talks to Buy AI Startup Decart for Around $7 Billion", 11 October 2026: the exclusive report of talks, approximate valuation, May round and valuation, chip-switching positioning and early Trainium3 access.
• Reuters via The Straits Times, 12 October 2026: confirmation that Reuters could not independently verify the Journal report; Lucy and Oasis product descriptions; May round with Nvidia; Anthropic's prior ~$6 billion exploration and September withdrawal.
• Bloomberg Law, 11 October 2026: corroboration of the Journal report and Anthropic's earlier diligence exit.
• CTech, 12 October 2026: chip-efficiency positioning, AWS customer relationship and sale-process history involving Nvidia then Anthropic.
• CTech earlier Trainium coverage and AI News on Lucy / Trainium3: background on Decart's Trainium performance claims, which are Decart-attributed.

No agreement has been announced. Valuation figures for a prospective deal rest on unnamed sources in the Journal's exclusive. Anthropic's reasons for withdrawing were not disclosed. The analysis and recommendations are ByteNib editorial analysis. This article was researched and drafted with AI assistance and edited under ByteNib's editorial standards.

Drafted with AI assistance and reviewed by the editor before publication.

Continue exploring: AI & Machine Learning analysis, the related implementation tutorial, and the structured learning path.